Hazard Assessment of Earthquake Disaster Chains Based on Deep Learning—A Case Study of Mao County, Sichuan Province

Abstract

Chain disasters often cause greater casualties and economic losses than single disasters. It plays an important role in the prevention and control to draw the susceptibility map and hazard map of geological hazards. To the best of our knowledge, the existing models are not suitable for the study of earthquake–geological disaster chains. Therefore, this study aims to establish a DNN model suitable for the study of earthquake–geological disaster chains. Firstly, nine key factors affecting geological disasters were selected and multi-source data sets were established based on geological disaster points in the study area. Secondly, the DNN model is trained to calculate the susceptibility of landslides and is discussed with the Support Vector Machine (SVM) model, Logistic Regression (LR) model, and Random Forest (RF) model. Finally, verify with the ROC curve. The verification results show that the DNN model has the highest accuracy among the proposed models. It is suitable for drawing geological hazard susceptibility maps and hazard maps. Therefore, it is proved that the model can be applied for the prediction of chain disasters and is a promising tool for geological hazard assessment.